Deep Fakes: Creation and detection techniques

Learn how a Deep Fake works and create videos with Face Swap and First Order Motion techniques, using practical and affordable tools.

What will you learn?

In this course, you will learn about the Deep Fake technique, which replaces the facial identity of people with others in videos. We will understand how Deep Fakes work from the techniques of Variational Autoencoders and GAN (Adversarial Generative Neural Network). Using Face Swap and First Order Motion techniques, we will produce videos from still images of faces. Along the way, we will learn to observe increasingly subtle Deep Fake signals and flaws, minimizing the chances of being fooled by them.

What is the course project?

Create medium-resolution Deep Fake videos based on static images and/or models trained on the faces of public figures, using Face Swap and First Order Motion techniques.


No prior technical knowledge is required. For this course, it is recommended that you have a computer or notebook with reasonable settings to run graphic applications and a good internet connection. It is not necessary to install any software beforehand.

  • AUDIO: Portuguese
  • SUBTITLES: English

Sergio Venancio

Sergio Venancio is a master and doctoral candidate in Visual Arts at ECA USP, Bachelor in Computer Science and in Visual Arts at UNICAMP. Professor at the INTELI Institute of Technology and Leadership, in the Graphic Design Specialization course – IA UNICAMP and in the Post-Graduation in Digital Architecture and Parametric Projects at Belas Artes SP. His research relates to the intersection between Visual Arts and Artificial Intelligences, experimenting with new forms of interactivity based on computer vision models, and producing software that simulates artistic actions, such as observational drawing.